Lecture 2: Motivating Example
Yale University
Consider historical trends in labor force participation (LFP) and earnings of women over 20th century.
Why did these trends occur?
Consider historical trends in labor force participation (LFP) and earnings of women over 20th century.
Why did these trends occur?
We will return to consider relevant econometric methods throughout the course.
Also motivated by Economists Amicus Brief in Dobbs v. Jackson Women’s Health
See Blau and Kahn (2017) for recent literature review.
Descriptive Statistics
Economic Models
Interpreting Changes in Measured Wage Gap
Determining Causality
Goldin (2006a) presents description statistics on women’s labor market outcomes.
The earnings gap between men and women, among those employed full-time, shrank dramatically starting around 1980.
Dramatic increase of women attending professional schools starting around 1970.
Goldin (2005) documents dramatic change in college majors for women starting around 1970, switching towards business and other traditionally male-intensive majors.
Descriptive Statistics
Economic Models
Interpreting Changes in Measured Wage Gap
Determining Causality
Why did these dramatic changes happen?
Starting point: Economic models of labor supply, wages.
Labor supply models:
substitution and income effects for labor supply,
role of fertility? marriage?
decision making within the household?
Why did these dramatic changes happen?
Starting point: Economic models of labor supply, wages.
Labor supply models:
substitution and income effects for labor supply,
role of fertility? marriage?
decision making within the household?
Why did these dramatic changes happen?
Starting point: Economic models of labor supply, wages.
Wages:
human capital model of wages
(education, work-experience),
discrimination, either “taste-based” or “statistical”
Why did these dramatic changes happen?
Starting point: Economic models of labor supply, wages.
Wages:
human capital model of wages
(education, work-experience),
discrimination, either “taste-based” or “statistical”
Dynamic models
Current wage and LFP depends on past choices
(past work experience, education, fertility…),
those past choices depend on expectations, e.g. education choices depend on expectations of fertility, LFP, etc.
LFP and earnings of women is a topic for economics.
Models of LFP and earnings naturally leads to studying:
fertility,
marriage,
divorce,
within household decision making/bargaining,
education decisions,
discrimination,
political economy. . .
Economists use economic models and econometrics to study all of these topics.
Descriptive Statistics
Economic Models
Interpreting Changes in Measured Wage Gap
Determining Causality
Could trends in wages among working women be due in part to changes in their observable characteristics, e.g., more women going to college? more work experience?
Use regression adjustment to study,
Selection into labor force:
we only observe wages of women who work.
Women who work might differ in unobserved ways from women who don’t work.
could trends in wages among working women be driven in part by which women work?
we will return to talk about selection later in this course.
Descriptive Statistics
Economic Models
Interpreting Changes in Measured Wage Gap
Determining Causality
Endogeneity
Fertility, education, labor force participation, etc,
jointly determined.
Complicates separating causation from correlation.
Endogeneity
Fertility, education, labor force participation, etc,
jointly determined.
Complicates separating causation from correlation.
relevant methodologies include natural-experiment approaches such as instrumental variables, we will study later in this course.
Timing: we can try to relate trends to events, e.g.,
Access to birth control pill: Griswold v. Connecticut 1965, Eisenstadt v. Baird 1971;
Access to abortion: Roe v. Wade 1973;
social movements and changes in social normals, e.g.,
2nd wave feminism in 1960s, 1970s.
How to use timing when so many events happened in same period?
Econometrics, natural-experiment approach: event study
combine across-time and across-group variation,
e.g., diff-in-diff and triple-diff.
Works well for some events, not for others:
Economics definition of discrimination:
Complicating issue: Relevant worker characteristics today can depend on past discrimination against worker or on worker’s expectations of future discrimination.
Economics definition of discrimination:
Same concept used for discrimination in loans, use of police force, judicial decisions (e.g., bail, sentencing, parole), etc., with same complicating issue.
How to measure/detect discrimination?
Can use regression adjustment for characteristics the researcher observes in the data.
Limitation:
discrimination is based on what employer observes, not on what we observe in the data.
problem of omitted variable bias, we will discuss extensively in this class.
How to measure/detect discrimination?
Natural experiment?
Goldin and Rouse (2000): Blind auditions
Most major orchestras switched to blind auditions in 1970s, 1980s
blind audition has candidates audition behind a screen, so that sex of candidate unknown to jury.
How to measure/detect discrimination?
How to measure/detect discrimination?
Traditional audit studies
Use pairs of actors, one minority and one non-minority, with similar relevant attributes.
Send a pair of actors to apply for each job (or loan, etc)
See if differences in outcomes for minority vs non-minority actors.
How to measure/detect discrimination?
Example: Neumark, Bank, and Van Nort (1996) discrimination in restaurant hiring.
Advantages of audit studies? limitations?
How to measure/detect discrimination?
Actual experiment?
Correspondence studies:
Create fictitious resumes, and randomly assign gender/minority status as signaled by name of applicant.
Send pairs of fictitious resumes to job or rental advertisements.
Measure difference in outcomes (typically call-backs) for applications with minority vs. non-minority names.
How to measure/detect discrimination?
Actual experiment?
Correspondence studies:
Famous example: Bertrand and Mullainathan (2004)
Advantages of correspondence studies? limitations?
cover regression analysis, instrumental variables, event studies, etc., in depth,
cover various applications, including related to LFP/fertility and to discrimination.
Econ 123: Lecture 2